Abstract Cancer is one of the leading causes of death in the United States, accounting for near 1 in every 4 deaths. However, despite the recent development of subtype-specific personalized therapy based on achievements in the fields of molecular and genetic profiling, many cancer treatments still have low efficacy which mostly arise from the limited ability to predict the patient tumor responses to therapeutic agents. The major reason that current therapeutics often cannot translate into a successful clinical outcomes is because of the complex tumor microenvironment and heterogeneity that limit the predictive power of the biomarker-guided strategies for chemotherapy. Therefore, the successful engineering of personalized three-dimensional (3D) tumor ecosystem that can recapitulate the tumor microenvironment and heterogeneity in vitro is strongly desired to accurately predict patients? responses to anti-cancer drugs and thus further improve patient outcome. Here we propose to develop a personalized breast-cancer-ecosystem-on-a-chip platform for personalized screening of cancer chemotherapeutics with high accuracy by utilizing patient-derived tumor explant, defined tumor grade-matched biomaterial matrices and autologous patient serum to mimic patient-specific tumor hallmarks. The proposed cancer-ecosystem-on-a-chip will also be tightly regulated under physiological fluid dynamics. In this project, we have hypothesized that 1) the use of tumor explant will embrace the critical components of the tumor heterogeneity of the patient, 2) the combination of defined tumor grade-matched matrix, autologous patient serum, and a microfluidic bioreactor will prevent the phenotype alteration of the tumor explant, and 3) the integration of a machine-learning algorithm with the cancer-ecosystem-on-a-chip platform will provide more accurate, unbiased prediction of the patient responses to chemotherapeutics based on the data gathered from the engineered tumor model. Our preliminary results show that the combination of tumor explant culture and tumor-derived matrix constituents had predicted therapeutic responses with 100% sensitivity. Our preliminary results show high specificity throughout a range of cancers including breast cancer, colorectal cancer, and head and neck squamous cell carcinoma, and thus the findings can have broad applications, and can emerge as a paradigm shift in the management of cancer.